remove max emd

Author: ghar1821

src/metrics/emd/config.vsh.yaml

@@ -55,50 +55,6 @@ info:
       max: .inf
       # Whether a higher value represents a 'better' solution (required)
       maximize: false
-      
-      # A unique identifier for your metric (required).
-      # Can contain only lowercase letters or underscores.
-    - name: emd_max_ct_horiz
-      # A relatively short label, used when rendering visualisarions (required)
-      label: EMD Max CT Horizontal
-      # A one sentence summary of how this metric works (required). Used when 
-      # rendering summary tables.
-      summary: "Max Earth Mover Distance calculated horizontally across donors for each cell type and marker."
-      # A multi-line description of how this component works (required). Used
-      # when rendering reference documentation.
-      description: |
-        Earth Mover Distance (EMD), also known as the Wasserstein metric, measures the difference 
-        between two probability distributions. 
-        
-        Here, EMD is used to compare marker expression distributions between paired samples from the same donor 
-        quantified across two different batches. 
-        For each paired sample, cell type, and marker, the marker expression values are first converted into 
-        probability distributions. 
-        This is done by binning the expression values into a range from -100 to 100 with a bin width of 0.1.
-        The `wasserstein_distance` function from SciPy is then used to calculate the EMD between the two 
-        probability distributions belonging to the same cell type, marker, and a given paired samples.
-        This is then repeated for every cell type, marker, and paired sample.
-        Finally, the maximum of all these EMD values is computed and reported as the metric score.
-
-        EMD Max CT score reflects the largest difference in marker expression distributions across all cell types, 
-        markers, and paired samples.
-        A high score indicates that at least one marker, cell type, or sample pair has a large difference in 
-        distribution after batch integration.
-        A low score means that even the most poorly corrected marker expression is well integrated across batches.    
-      references:
-        doi: 
-          - 10.1023/A:1026543900054
-      links:
-        # URL to the documentation for this metric (required).
-        documentation: https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.wasserstein_distance.html
-        # URL to the code repository for this metric (required).
-        repository: https://github.com/scipy/scipy
-      # The minimum possible value for this metric (required)
-      min: 0
-      # The maximum possible value for this metric (required)
-      max: .inf
-      # Whether a higher value represents a 'better' solution (required)
-      maximize: false
 
       # A unique identifier for your metric (required).
       # Can contain only lowercase letters or underscores.
@@ -140,47 +96,6 @@ info:
       max: .inf
       # Whether a higher value represents a 'better' solution (required)
       maximize: false
-      
-      # A unique identifier for your metric (required).
-      # Can contain only lowercase letters or underscores.
-    - name: emd_max_ct_vert
-      # A relatively short label, used when rendering visualisarions (required)
-      label: EMD Max CT Vertical
-      # A one sentence summary of how this metric works (required). Used when 
-      # rendering summary tables.
-      summary: "Max Earth Mover Distance across batch corrected samples, cell types, and markers."
-      # A multi-line description of how this component works (required). Used
-      # when rendering reference documentation.
-      description: |
-        Earth Mover Distance (EMD), also known as the Wasserstein metric, measures the difference 
-        between two probability distributions. 
-        
-        Here, EMD is used to compare marker expression distributions between all integrated 
-        samples from the same group.
-        For each pair of samples, cell type, and marker, the marker expression values are first converted into 
-        probability distributions. 
-        This is done by binning the expression values into a range from -100 to 100 with a bin width of 0.1.
-        The `wasserstein_distance` function from SciPy is then used to calculate the EMD between the two 
-        probability distributions belonging to the same cell type, marker, and a given paired samples.
-        This is then repeated for every cell type, marker, and paired sample.
-        Finally, the maximum of all these EMD values is computed and reported as the metric score.
-
-        A high score indicates there is a pair of samples and marker which show large difference in distribution after batch integration.
-        A low score means that, the worst integrated pair of samples and marker are well integrated.
-      references:
-        doi: 
-          - 10.1023/A:1026543900054
-      links:
-        # URL to the documentation for this metric (required).
-        documentation: https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.wasserstein_distance.html
-        # URL to the code repository for this metric (required).
-        repository: https://github.com/scipy/scipy
-      # The minimum possible value for this metric (required)
-      min: 0
-      # The maximum possible value for this metric (required)
-      max: .inf
-      # Whether a higher value represents a 'better' solution (required)
-      maximize: false
 
 # Resources required to run the component
 resources:
@@ -203,4 +118,4 @@ runners:
   # Allows turning the component into a Nextflow module / pipeline.
   - type: nextflow
     directives:
-      label: [midtime,midmem,midcpu]
+      label: [lowtime,lowmem,lowcpu]

src/metrics/emd/helper.py

@@ -6,7 +6,6 @@
 from scipy.stats import wasserstein_distance
 
 KEY_MEAN_EMD_CT = "mean_emd_ct"
-KEY_MAX_EMD_CT = "max_emd_ct"
 KEY_EMD_VERT_MAT_split1 = "emd_vert_mat_split1"
 KEY_EMD_VERT_MAT_split2 = "emd_vert_mat_split2"
 KEY_EMD_HORZ_PER_DONOR = "emd_horz_per_donor"
@@ -27,13 +26,14 @@ def calculate_vertical_emd(
         dict: a dictionary containing the following elements.
             "mean_emd_ct": np.float32: mean emd value computed from a flattened data frame containing
                 mean emd computed for every marker and cell type across all pairing two samples from the same group.
-            "max_emd_ct": np.float32: max emd value computed from a flattened data frame containing
-                max emd computed for every marker and cell type across all pairing two samples from the same group.
-            "emd_wide_dfs": dict: each key is a marker. A value is yet another dictionary which value is
-                a 2d matrix where each row/column is a pair of sample and a cell contains
-                    emd value computed for the sample pair for either a given cell type
-                    or global. The key for this dictionary then is either a given cell type
-                    or global, depending on what the 2d matrix represents.
+            "emd_split1_long": pd.DataFrame or np.nan: a long format dataframe where each row is a combination of
+                first_sample, second_sample, cell_type, and markers.
+                Value is the EMD computed for that combination.
+                If the input data does not have at least 2 samples per group, then return np.nan.
+            "emd_split2_long": pd.DataFrame or np.nan: a long format dataframe where each row is a combination of
+                first_sample, second_sample, cell_type, and markers.
+                Value is the EMD computed for that combination.
+                If the input data does not have at least 2 samples per group, then return np.nan.
     """
 
     print("Calculating vertical EMD for split 1", flush=True)
@@ -48,9 +48,8 @@ def calculate_vertical_emd(
 
     # safeguard
     mean_emd_ct = np.nan
-    max_emd_ct = np.nan
 
-    print("Computing mean and max vertical EMD", flush=True)
+    print("Computing mean vertical EMD", flush=True)
 
     # compute these only if we can.
     emd_long = []
@@ -71,15 +70,9 @@ def calculate_vertical_emd(
             .to_numpy()
             .flatten()
         )
-        max_emd_ct = np.nanmax(
-            emd_long.drop(columns=["cell_type", "first_sample", "second_sample"])
-            .to_numpy()
-            .flatten()
-        )
 
     return {
         KEY_MEAN_EMD_CT: mean_emd_ct,
-        KEY_MAX_EMD_CT: max_emd_ct,
         KEY_EMD_VERT_MAT_split1: emd_split1_long,
         KEY_EMD_VERT_MAT_split2: emd_split2_long,
     }
@@ -94,9 +87,10 @@ def get_vert_emd_for_integrated_adata(i_adata: ad.AnnData, markers_to_assess: li
         markers_to_assess (list): list of markers to compute EMD for.
 
     Returns:
-        emd_dfs_ct: EMD per cell type,
-        emd_dfs_global: EMD per donor,
-        emd_wide_dfs: the break down of EMD per donor and cell type.
+        emd_vals (pd.DataFrame or np.nan): a long format dataframe where each row is a combination of
+            first_sample, second_sample, cell_type, and markers.
+            Value is the EMD computed for that combination.
+            If the input data does not have at least 2 samples per group, then return np.nan.
     """
 
     print("Determining samples per group", flush=True)
@@ -278,9 +272,6 @@ def calculate_horizontal_emd(
     mean_emd_ct = np.nanmean(
         emd_per_donor_per_ct.drop(columns=["cell_type", "donor"]).values
     )
-    max_emd_ct = np.nanmax(
-        emd_per_donor_per_ct.drop(columns=["cell_type", "donor"]).values
-    )
 
     # concatenate the global and cell type emd
     emd_per_donor_per_ct.columns = emd_per_donor_per_ct.columns.str.replace(
@@ -289,7 +280,6 @@ def calculate_horizontal_emd(
 
     return {
         KEY_MEAN_EMD_CT: mean_emd_ct,
-        KEY_MAX_EMD_CT: max_emd_ct,
         KEY_EMD_HORZ_PER_DONOR: emd_per_donor_per_ct,
     }
 

src/metrics/emd/script.py

@@ -99,15 +99,11 @@
         "method_id": method_id,
         "metric_ids": [
             "emd_mean_ct_horiz",
-            "emd_max_ct_horiz",
             "emd_mean_ct_vert",
-            "emd_max_ct_vert",
         ],
         "metric_values": [
             emd_horz[emd_helper.KEY_MEAN_EMD_CT],
-            emd_horz[emd_helper.KEY_MAX_EMD_CT],
             emd_vert[emd_helper.KEY_MEAN_EMD_CT],
-            emd_vert[emd_helper.KEY_MAX_EMD_CT],
         ],
         "emd_values": {
             "emd_values_horiz": emd_horz[emd_helper.KEY_EMD_HORZ_PER_DONOR],